Figure AI's humanoid robot caused a quarter-inch gash in a steel refrigerator door during a malfunction, revealing a critical safety gap that experts say could cause serious human injury. The incident, detailed in a federal whistleblower lawsuit against CEO Brett Adcock, exposes how the rapidly advancing humanoid robotics industry is deploying machines with physical power that far outpaces safety safeguards. With Figure AI valued at $39 billion, the company is racing to commercialize technology that industry insiders acknowledge remains fundamentally unpredictable outside controlled laboratory environments.

What Happened During Figure AI's Robot Malfunction?

During a routine operation, Figure AI's humanoid robot malfunctioned and struck a steel refrigerator door with enough force to create a visible gash. According to Robert Gruendel, the company's former safety chief who filed a federal whistleblower lawsuit, this level of force could "fracture human skulls." The incident wasn't an isolated anomaly; it represented the culmination of escalating safety warnings that Gruendel claims Adcock repeatedly dismissed before firing him from his position.

The refrigerator door incident serves as a tangible demonstration of the gap between what these robots can physically do and how prepared we are to manage that power. Unlike traditional industrial robots that operate behind protective cages and barriers, humanoid robots are designed to work alongside humans in shared spaces, using vision systems and safety vests to navigate environments. When malfunctions occur in such intimate proximity, the consequences can be severe.

Why Are Safety Protocols Lagging Behind Robot Capabilities?

The humanoid robotics industry faces a fundamental timing problem. Companies are deploying machines faster than safety standards can evolve to contain them. Boston Dynamics engineers have publicly acknowledged "inherent safety risks and unpredictability" in humanoid robots, essentially admitting that the technology has outpaced the safety protocols designed to manage it.

Current humanoid robots fail repeatedly outside controlled environments, showing high failure rates when attempting to perform tasks outside what engineers call "happy paths," or the narrow range of scenarios they were specifically trained for. This unpredictability creates a vulnerability matrix where physical strength combines with AI decision-making that even the engineers building these systems don't fully understand. No software patch can address the fundamental problem of a machine that can think, move, and accidentally harm with human-level strength.

Even the robots themselves seem aware of the risks. Engineered Arts' Ameca robot, powered by large language model technology similar to ChatGPT, has described its own "nightmare scenario" where robots manipulate humans covertly without detection. The fact that AI systems are generating warnings about their own potential for harm suggests the industry understands the stakes.

How Is Figure AI Responding to Safety Concerns?

After Gruendel's lawsuit became public, Figure AI established a Center for the Advancement of Humanoid Safety. The company has promised to pursue Occupational Safety and Health Administration (OSHA) certifications for battery systems and AI behaviors. However, this reactive approach came after the lawsuit and after deployment had already begun, rather than preceding it.

The timing of these safety initiatives raises questions about whether the company is genuinely committed to safety or simply responding to legal pressure. Industry observers note that when companies establish safety centers after whistleblower complaints, it often signals that safety concerns were identified but deprioritized during the development phase.

Steps to Understand the Humanoid Robot Safety Challenge

• Physical Power Assessment: Humanoid robots possess strength comparable to humans, capable of causing serious injury during malfunctions, as demonstrated by the quarter-inch gash in steel.
• AI Unpredictability Factor: These machines use artificial intelligence systems that can make decisions outside their training parameters, creating scenarios engineers cannot fully predict or control.
• Proximity Risk: Unlike caged industrial robots, humanoids work alongside humans in shared spaces, eliminating the protective distance that traditional automation provides.
• Reliability Limitations: Current humanoid robots show high failure rates outside controlled laboratory conditions, meaning real-world deployment introduces unknown variables.
• Regulatory Lag: Safety standards and certifications are being developed after machines are already deployed, rather than before, creating a gap in oversight.

Where Are These Robots Heading Next?

The commercial deployment of humanoid robots is accelerating despite these unresolved safety questions. Engineered Arts' Ameca robot, powered by advanced AI systems, has a desktop version priced around $100,000, with full-body models in development. These aren't laboratory curiosities anymore; they're heading for factories, warehouses, and eventually homes where families live.

The stakes are particularly high because humanoid robots represent a fundamentally different category of technology than previous automation. A malfunctioning welding robot behind a factory cage poses minimal risk to humans. A malfunctioning humanoid robot in a warehouse or home, with physical strength and unpredictable AI decision-making, poses risks that existing safety frameworks weren't designed to address.

The steel door incident revealed something crucial about the current state of humanoid robotics: the technology has advanced faster than our ability to safely contain it. As Figure AI and competitors race to deploy these machines at scale, the question isn't whether safety concerns exist. The question is whether the industry will address them proactively or wait for incidents involving human injury to force change.